[pymvpa] Combinatorial MVPA

[Yaroslav Halchenko]

On Wed, 02 Dec 2015, Bill Broderick wrote:

> Hi all,

> I'm trying to put together an analysis using "combinatorial MVPA", and I'm not
> sure the best way to do it in PyMVPA.

> Instead of using voxel intensities or voxel-wise betas, our input data is 26
> time courses, one each for 25 networks and one for an ROI. For each time
> course, we have a volume for each volume, so creating the dataset and running
> regular MVPA is pretty straightforward.

could you please share ds.summary() for the dataset you have
constructed to get a better picture of "a volume for each volume" ;)

> However, to determine which timecourse is contributing the most to the
> classifiers performance,

there is yet another black hole of methods to assess contribution of
each feature to performance of the classifier.  The irelief, which was
mentioned is one of them... 

But! in all of those cases, I would say, that you should not draw any
strong conclusions if you are not getting good classification
performance.  So what is your classification performance if you just do
classsification on all features?  which one could you obtain if you do
feature selection, e.g. with SplitRFE (which would eliminate features to
attain best performance within each cv folds in nested cv)
https://github.com/PyMVPA/PyMVPA/blob/master/mvpa2/featsel/rfe.py#L415
-- 
Yaroslav O. Halchenko
Center for Open Neuroscience     http://centerforopenneuroscience.org
Dartmouth College, 419 Moore Hall, Hinman Box 6207, Hanover, NH 03755
Phone: +1 (603) 646-9834                       Fax: +1 (603) 646-1419
WWW:   http://www.linkedin.com/in/yarik